. . Polity Volume 46, Number 3 July 2014 © 2014 Northeastern Political Science Association 0032-3497/14 www.palgrave-journals.com/polity/ A More Responsible Two-Party System? Accountability for Majority and Minority Party Performance in a Polarized Congress David R. Jones Baruch College and the Graduate Center, City University of New York O R C O PY Existing research suggests that to the extent that accountability for congressional performance occurs, it does so primarily through a referendum on the performance of the majority party. If true, this means that the minority party has no incentive to act responsibly, and may even have an incentive to polarize and obstruct. I argue that Americans do not simply hold one summary evaluation of congressional performance. Instead, partisan polarization in Congress has allowed Americans to form distinct evaluations of each congressional party. When casting ballots in House elections, Americans take into account their evaluations of both congressional parties. As a result, American politics is now closer to the “responsible two-party system” ideal set by the American Political Science Association more than six decades ago. Polity (2014) 46, 470–492. doi:10.1057/pol.2014.12 TH Keywords polarization; congress; political parties; accountability; elections; approval ratings A U The opposition is as responsible for its record in Congress as is the party in power. It is important that the opposition party be effective but it is equally important that it be responsible, for an irresponsible opposition is dangerous to the whole political system. American Political Science Association, 1950 In 2009, as House Republican leaders prepared for the start of the 111th Congress, they privately decided their party should “show united and unyielding opposition” to the economic agenda of President Obama and the Democratic congressional majority.1 The reasons for this decision were not policy oriented, but rather explicitly electoral: a PowerPoint prepared by the House leadership stated that while the purpose of the majority is to govern, “the purpose of the minority is to 1. Robert Draper, Do Not Ask What Good We Do: Inside the U.S. House of Representatives (New York: Free Press, 2012), xix. David R. Jones 471 A U TH O R C O PY become the majority.”2 This strategy was not entirely new. In 2005, Democrats facing a Republican majority in Congress and a Republican President adopted one that was strikingly similar.3 In both cases, the minority party appeared to believe that it could make the ruling party look bad by denying it bipartisan legislative victories, and that the public would respond at the polls by punishing the ruling party, not the minority. Also in both cases, the strategy seemed to pay dividends. The minority party experienced sweeping electoral successes two years later, regaining their majority status. From a normative perspective, the examples presented above are quite troubling for American democracy. First, they suggest that the majority party may be held electorally accountable for a legislative performance that is partly due to the actions of the minority party. Second, they suggest that the minority party in Congress is not held electorally accountable for its actions. Assuming both of these things are true, the minority party would appear to have every electoral incentive to polarize Congress along partisan lines and otherwise obstruct legislative business. Finally, the notion that minority parties may have a strong incentive to act “irresponsibly” is particularly troubling in an era of split party control of Congress. In periods when Democrats are a minority in one chamber and Republicans the minority in the other—such as 2011–2014—the electoral incentive to polarize and obstruct may be felt by both parties in their respective minority chambers. In sum, our system of government may be structured to reward irresponsible behavior in Congress and encourage further polarization of politics. While some scholars have theorized that the majority party’s behavior in office is driven in part by its concern over how that performance will be evaluated by the public,4 very little has been theorized about whether the minority party is or should be concerned about its performance in Congress. Furthermore, empirical literature on the causes and effects of performance ratings in Congress has focused predominantly on the performance ratings of Congress as a collectivity (e.g., “ … the job done by Congress”) rather than explicitly considering the performance ratings of each congressional party individually (e.g., “ … the job done by Republicans in Congress”).5 As a result, the theories we do have regarding 2. Michael Grunwald, The New New Deal: The Hidden Story of Change in the Obama Era (New York: Simon & Schuster, 2012), 143. 3. Ronald Brownstein, “Democrats Seek to Outmaneuver Republicans by Imitating Their Strategy,” Los Angeles Times, February 14, 2005. 4. Gary W. Cox and Matthew D. McCubbins, Setting the Agenda: Responsible Party Government in the U.S. House of Representatives (New York: Cambridge University Press, 2005). 5. For example, John R. Hibbing and Elizabeth Theiss-Morse, Congress as Public Enemy: Public Attitudes towards American Political Institution (New York: Cambridge University Press, 1995); David R. Jones and Monika L. McDermott, Americans, Congress, and Democratic Responsiveness: Public Evaluations of Congress and Electoral Consequences (Ann Arbor: The University of Michigan Press, 2009); Mark Ramirez, “The Dynamics of Partisan Conflict on Congressional Approval,” American Journal of Political Science 53 (2009): 681–94. 472 A MORE RESPONSIBLE TWO-PARTY SYSTEM? performance-related incentives of the parties in Congress are largely untested with direct empirical evidence. This study examines the American public’s evaluations of each congressional party’s performance and the role of these evaluations in elections. I argue that the ratings Americans give to each congressional party flow in part from the parties’ actual performances in Congress, reflecting factors such as policy stances, managerial effectiveness, and partisan tone. I also argue that when Americans cast ballots in elections, they take into account their evaluations of not only majority party performance, but also their separate evaluations of minority party performance. In this way, Americans hold both parties accountable for their respective performances.6 PY General Theories of Accountability for Collective Performance A U TH O R C O It is possible to distinguish between two general types of electoral accountability for collective performance in government. The first type can be loosely characterized as a one-party referendum model. In this model, an individual typically forms a single impression of government’s overall performance. This impression is associated in the individual’s mind with one party: the “party in power.” If this overall impression is positive, the individual is more inclined to vote to keep that party in power. If this overall impression is negative, the individual is more inclined to vote against that party. In the literature, this model is close in nature to the work of V.O. Key, Jr.7 It also has a connection to literature arguing that public evaluations of government performance are largely “generic.”8 The central premise of this literature is that citizens do not form highly distinct evaluations of different actors in government. Instead, they are more likely to have a single impression of government performance. They may answer survey questions about the performance of a variety of actors in government, but in reality all answers revolve around the same basic impression. The second general type of accountability can be characterized as a two-party model. In this model, an individual forms an impression of each party based on the 6. The same cannot necessarily be said if Americans only take into account a single, collective evaluation of congressional performance. For example, minority party obstruction may lead to poor overall ratings of Congress, but the minority party is not held accountable for its obstruction if (as existing literature predicts) the electoral consequence is seat losses by the majority and corresponding gains by the minority. For further discussion, see the following section. 7. V. O. Key, Jr., The Responsible Electorate: Rationality in Presidential Voting, 1936–1960 (New York: Vintage Books, 1966). 8. Wendy M. Rahn, “Affect as Information: The Role of Public Mood in Political Reasoning,” in Elements of Reason: Cognition, Choice, and the Bounds of Rationality, ed. Arthur Lupia, Mathew D. McCubbins, and Samuel L. Popkin (New York: Cambridge University Press, 2000), 130–50; James Stimson, Tides of Consent: How Public Opinion Shapes American Politics (New York: Cambridge University Press, 2004). David R. Jones 473 A U TH O R C O PY performance of its respective members in office. In this case, unlike the one-party model, if an individual’s rating of the ruling party is negative, a voter will not necessarily tend to vote against that party. Instead, her vote will also be conditioned by her rating of the opposition party. If the opposition party’s rating makes it an acceptable alternative, then she may indeed vote against the ruling party. But even though the voting decision might look the same as in the one-party model, the implication is slightly different: it is not just a vote to blindly punish the ruling party; it is also a vote in favor of the opposition. The voter is taking into account evaluations of both parties, not just a single evaluation applied to the party in power. In the literature, this second model shares much in common with the normative “responsible party government” model, particularly as outlined in the 1950 American Political Science Association (APSA) report, Towards a More Responsible Two-Party System. As the report states, “Party responsibility means the responsibility of both parties [the party in power and the opposition party] to the general public, as enforced in elections.”9 This model is often associated with the idea that public evaluations of political actors are formed largely on the basis of policy stances, and has been criticized on the basis that many Americans do not appear to care about policy.10 However, a restriction to policy criteria is not a necessary component of the model. For example, the APSA report says that party responsibility also includes the parties’ “manner” and their “conduct,” as well as “results achieved.”11 Of course both of these models are complicated by the fact that in the United States’ system of separate legislative and executive elections, there is no guarantee that we will have one clear party in power. In particular, the frequent occurrence of divided government, as well as occasional friction between a president and his own party in Congress even during unified government, have necessitated additional theorizing about whether and how models of collective accountability should be more realistically adapted.12 Application of the One-Party Model to Congress At first, the answer to the question of how we identify the “party in power” in our separated system was to simply ignore Congress and focus on the presidency.13 9. American Political Science Association, Committee on Political Parties, Toward a More Responsible Two-Party System (New York: Rinehart, 1950), 22. 10. For example, see Hibbing and Theiss-Morse, Congress as Public Enemy, 2. 11. APSA, Toward a More Responsible Two-Party System, 22. 12. See, for example, James L. Sundquist, “Needed: A Political Theory for the New Era of Divided Government in the United States,” Political Science Quarterly 103 (1988): 613–35. 13. For example, see Key, Responsible Electorate, 74: “ … all the issues the voter faced in 1952 were bundled up in the question of whether he approved of the performance of the party in power. And that became in reality a question of whether he approved on Mr. Truman.” 474 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY Specifically, when citizens disapprove of the way the current occupant of the White House is handling his job as president, they are less likely to vote for his reelection, or for his party’s nominee for president, or for his party’s candidates for Congress.14 In contrast to the prominent role in elections accorded to presidential party performance, early research suggested almost no role for congressional party performance. More recently, McDermott and Jones have developed and tested an argument regarding congressional performance.15 Consistent with the notion that citizens focus on the “party in power,” they hypothesize that Americans view Congress as being controlled by the majority party, and therefore a voter’s approval of congressional performance will increase her likelihood of casting a House ballot for the majority party’s candidate. Using ANES data on House elections from 1980 to 2000, they find support for this hypothesis not only in races with majority party incumbents, but also in open seat races and in races with minority party incumbents. Subsequent research also finds support for this hypothesis in Senate voting,16 in aggregate election results,17 and in even in elections from the 1950s, 1960s and 1970s.18 In addition, the effect of congressional performance evaluations on majority party fortunes appears to be growing as Congress polarizes along party lines.19 On its face, these findings would appear to support the one-party model of accountability for Congress. Specifically, they are consistent with a story in which everything a citizen learns or feels about Congress is automatically associated with a single party: the majority. But while it may be true that Americans closely associate approval of Congress with approval of the majority party, this does not rule out the possibility that Americans also evaluate the performance of the minority party in Congress and use that evaluation when voting. No existing studies offer or test this alternative hypothesis. 14. For example, see Morris P. Fiorina, Retrospective Voting in American National Elections (New Haven: Yale University Press, 1981). 15. Monika L. McDermott and David R. Jones, “Do Public Evaluations of Congress Matter? Retrospective Voting in Congressional Elections,” American Politics Research 32 (2003): 155–77. 16. David R. Jones and Monika L. McDermott, “The Responsible Party Government Model in House and Senate Elections,” American Journal of Political Science 48 (2004): 1–12; Monika L. McDermott and David R. Jones, “Congressional Performance, Incumbent Behavior, and Voting in Senate Elections,” Legislative Studies Quarterly 30 (2005): 235–57. 17. Jones and McDermott, Americans, Congress, and Democratic Responsiveness; David R. Jones and Monika L. McDermott, “The Salience of the Democratic Congress and the 2010 Elections,” PS: Political Science and Politics 44 (2011): 297–301. 18. David R. Jones, “Evaluations of Congress and Voting in House Elections: Revisiting the Historical Record,” Public Opinion Quarterly 74 (2010): 696–710. 19. David R. Jones, “Partisan Polarization and Congressional Accountability in House Elections,” American Journal of Political Science 54 (2010): 323–37. David R. Jones 475 Application of the Two-Party Model to Congress A U TH O R C O PY In contrast to previous research, this study argues that Congress is held accountable for its performance not solely through a referendum on the performance of the majority party. Rather, accountability for congressional performance shares much in common with the two-party model. In this section I sketch out how a twoparty theory might apply specifically to Congress. What goes on in Congress provides potentially useful information about the fitness for office not just of the majority party but also of the minority party. As the authors of the APSA report note, “In Congress, the two parties are on almost daily trial … [Party] reputations … are made or lost.”20 In particular, it is possible to identify three general types of information about the congressional parties that can be gleaned from their performances.21 First, party performance in Congress can reveal information about a party’s professional competence in getting things done. This type of information may apply mainly to the majority not the minority party.22 As Cox and McCubbins argue, “The more favorable the majority party’s record of legislative accomplishment, the better its reputation or brand name will be.”23 Existing research presents no corollary argument suggesting that legislative accomplishment will affect the reputation of the minority party. This is logical, since a party that does not have enough votes to pass anything on its own would appear to bear little responsibility for the level of legislative achievement, particularly in comparison to a party that comprises a majority of votes and occupies all the key leadership positions.24 A second type of information that can be gleaned from party performance in Congress is a party’s policy stances. In contrast to information about legislative accomplishment, this type of information is available for both parties. When speeches are made and votes are cast in Congress, each congressional party essentially registers its level of agreement or disagreement with a policy. Citizens can then use this information to form a general sense of where each congressional party stands on the issues in relation to their own beliefs. This idea is closely related 20. APSA, Toward a More Responsible Two-Party System, 56. 21. While the analysis controls for economic conditions, I choose not to treat this variable as related to congressional performance for two reasons. From a theoretical perspective, the potential link between congressional actions and the economy is, at best, much more indirect than the link with the other criteria listed here. And empirically, research shows that the public associates economic performance with the president, not Congress. 22. This means that during split control congresses, it would apply to the respective majority in each chamber. 23. Cox and McCubbins, Setting the Agenda, 7. 24. Filibusters by the Senate minority party would be an important exception. See the concluding section for additional discussion of this point. 476 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY to the literature on party ideological brand names emanating from voting patterns in Congress.25 A third type of information available from observing congressional party performance is the manner in which the parties go about the business of Congress. More specifically, the parties may proceed in a cooperative, bipartisan fashion or they may act more conflictually.26 Again, unlike legislative accomplishment, this information could potentially be relevant to each party. The majority party plays a role in setting the partisan tone of Congress in part by deciding how inclusive or exclusive the legislative process will be. The minority party plays a role when it decides whether it will work constructively with the majority or staunchly oppose it. In theory, one, both, or neither party might be responsible for the level of partisan conflict through their behavior in Congress. Just because these three types of information could theoretically be used by citizens does not mean that they are always used. Instead, the extent to which this information can be used is in large part a function of how internally cohesive each party in Congress is. When a party is not cohesive, it is less likely to project a clear image to citizens. On the other hand, when a party is cohesive, it projects a clearer image, and voters are more likely to be able to use that image as a heuristic to help them guess how other members of that party might act if elected. Indeed, the 1950 APSA report advocated for greater “party cohesion in Congress” because it said this would increase the ability of Americans to hold the parties accountable.27 Since the 1970s, the congressional parties have grown more cohesive and more distinct from one another—in other words, they have grown more polarized.28 As a result, the public is now more aware of the distinction between Democrats and Republicans in Congress.29 This has led scholars such as Barbara Sinclair to conclude that “political parties seem to meet the requirements of responsible parties as defined by [the 1950 APSA report] to a greater extent today than at any time in the past half century.”30 Since parties in Congress are more distinct and Americans better perceive this distinction, this should mean that citizens are now more likely to use information from both parties’ performances in Congress. 25. Jonathan Woon and Jeremy C. Pope, “Made in Congress? Testing the Electoral Implications of Party Ideological Brand Names,” The Journal of Politics 70 (2008): 823–36. 26. Hibbing and Theiss-Morse, Congress as Public Enemy; Ramirez, “Dynamics of Partisan Conflict.” 27. APSA, Toward a More Responsible Two-Party System, 21. 28. For example, Nolan McCarty, Keith Poole, and Howard Rosenthal, Polarized America: The Dance of Unequal Riches (Cambridge, MA: MIT Press, 2006). 29. Marc J. Hetherington, “Resurgent Mass Partisanship: The Role of Elite Polarization,” American Political Science Review 95 (2001): 619–32; Jeremy C. Pope and Jonathan Woon, “Measuring Changes in American Party Reputations, 1939–2004,” Political Research Quarterly 62 (2009): 653–61. 30. Barbara Sinclair, Party Wars: Polarization and the Politics of National Policy Making (Norman: University of Oklahoma Press, 2006). David R. Jones 477 PY In particular, citizens might use this information in elections. In the past, when it was difficult to develop firm views of each party’s particular fitness for office based on its performance in Congress, citizens may have been more likely to default to one-party referendum mode, issuing an up or down verdict to whichever party happened to be nominally in control of a complicated Congress. But with more distinct performance-based information about both parties, citizens have an increased ability to also consider whether one party might be better or worse than the other party. Given the option of both modes of voting, it would seem logical for a voter to use the two-party model of accountability when parties are distinct rather than the one-party model. Otherwise, a voter risks returning to power a party that is less-preferred when all relevant information is considered. Put another way, it would not make sense for a voter to “throw away” information that can improve the quality of a voting decision.31 C O Analysis of How Americans Evaluate Congressional Parties A U TH O R The goal of this study is to show that the public holds both parties accountable for their respective performances in Congress. True accountability requires that citizens form distinct ratings of each party’s performance in Congress, and that they use the performance-related criteria outlined in the previous section when forming these evaluations. If both congressional parties receive essentially the same rating, then each party is not being evaluated on its unique performance. If the evaluations are different from each other but still unrelated to performancebased criteria (e.g., purely a function of party identification), then even if these evaluations are correlated with voting, the congressional parties are not being held accountable for their actual performance. To test whether Americans form distinct evaluations of each party related to performance, I employ the 2006 Congressional Elections Study (CES), a survey conducted by the Center for Survey Research at Indiana University.32 I use this survey because unlike standard political science surveys (American National Election Study, Cooperative Congressional Election Study) it specifically asks respondents for their evaluations of “congressional Democrats” and of “congressional Republicans,” enabling an analysis of the two-party model. In addition, it includes other non-standard questions that approximate the performance criteria hypothesized to influence these evaluations, including a respondent’s assessment 31. Another reason that voters might follow a two-party model rather than a one-party model is that while the latter requires a voter to know which party controls congress (a question increasingly complicated by split-party control of each chamber), the former does not. For more on knowledge of party control of Congress, see Jones and McDermott, Americans, Congress, and Democratic Responsiveness, 46-49. 32. Additional details on this survey and coding are included in the Appendix. 478 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY of the degree of legislative productivity, the degree of legislative partisanship, and the ideological stance of each congressional party. The analysis also controls for other variables commonly believed to influence performance ratings, including a respondent’s party identification, rating of the national economy, feelings of external efficacy, and rating of one’s own incumbent. The first question in this analysis is whether Americans give the same rating to both parties in Congress, or whether their party ratings are distinct. In the 2006 CES, the aggregate ratings of the two congressional parties appear somewhat similar. Congressional Republicans are rated negatively by 49 percent of respondents and positively by 19 percent with the remainder neutral, while congressional Democrats are rated negatively by 41 percent and positively by 20 percent with the remainder neutral. However, these aggregate figures mask considerable differences in individuals’ ratings of each party: among respondents who rate Republicans in Congress negatively, only one-third rate Democrats in Congress the same way; among respondents who rate Democrats in Congress negatively, only two-fifths rate Republicans in Congress the same way. In fact, contrary to the idea of generic ratings of the performance of everyone in government, individuals’ ratings of the two congressional parties are actually negatively correlated, not positively correlated, although this relationship is relatively weak (r = −0.19). Next, I analyze whether the ratings Americans give to each congressional party are related to performance-based criteria (as opposed, for example, to being purely based on party identification). Based on the discussion of performancebased criteria outlined in the previous section, I have three particular expectations: (1) greater concern about a lack of legislative productivity will lower evaluations of the congressional majority party but not affect evaluations of the minority party; (2) greater concern about the partisan tone in Congress will lower evaluations of both congressional parties (with the relative size of the effect for each party dependent on perceived responsibility); (3) greater divergence between a congressional party and a citizen in terms of policy preference will lower evaluations of that party’s performance in Congress. Table 1 presents the results of three ordinal regression analyses. As a point of comparison, the first model in the table estimates the impact of the independent variables on the more traditional measure of congressional performance: approval of the collective Congress. In the second model the dependent variable is evaluations of congressional Republicans (the majority in 2006), and in the third model it is evaluations of congressional Democrats (the minority in 2006). Each of the three ordinal regression models is followed by a column that translates the results into probabilistic terms. Specifically, the numbers in these “effects” columns represent the estimated increase in the probability of holding either a neutral or positive evaluation (as opposed to a negative evaluation) as one moves from the PY Table 1 Explaining Evaluations of Performance in Congress, 2006 U A Note: Number of cases is 862 for all models. *p<0.05 (one-tailed). O −0.42 −0.47 −0.12 −0.08 0.15 0.22 0.21 0.56 0.41 Coefficient −0.356* −0.364* −0.520* 0.247* 0.395* 0.365* 0.100* 0.268* 0.075 0.548 R (0.077) (0.076) (0.066) (0.066) (0.040) (0.071) (0.057) (0.085) (0.089) TH −0.465* −0.520* −0.139* −0.097 0.110* 0.245* 0.231* 0.714* 0.507* 0.444 Effect SE Effect (0.074) (0.072) (0.069) (0.063) (0.040) (0.070) (0.056) (0.079) (0.089) −0.34 −0.34 −0.47 0.24 0.53 0.35 0.10 0.26 0.07 Congressional Democrats (Minority) Coefficient −0.093 −0.250* 0.178* −0.540* −0.346* −0.193* 0.138* 0.000 0.173* 0.392 SE Effect (0.071) (0.069) (0.061) (0.063) (0.039) (0.067) (0.053) (0.074) (0.082) −0.09 −0.23 0.21 −0.49 −0.46 −0.19 0.13 0.00 0.17 David R. Jones Unproductive Congress Overly partisan Congress Policy divergence of Reps. Policy divergence of Dems. Party identification National economy External efficacy Rating of incumbent (Rep.) Rating of incumbent (Dem.) Nagelkerke R2 SE O Coefficient Congressional Republicans (Majority) C Collective congress 479 480 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY lowest category to the highest category of that variable, holding all other variables constant at their mean. The first model conforms to what previous research has already shown regarding evaluations of the collective Congress. Specifically, the results show that Americans dislike a Congress they perceive as unproductive and as overly partisan, and in which the party in charge (here, Republicans) poorly represents their policy views. The control variables also behave as expected, with positive ratings being given to Congress by those who identify with the majority party, who feel the economy is good, who feel efficacious, and who approve of their own incumbent. More importantly, the results show that Americans’ evaluations of congressional Republicans and congressional Democrats are not simply identical replications of their evaluations of the collective Congress. As expected, the first row in Table 1 shows that the public appears to apply the productivity criterion only to its evaluations of the majority party, not the minority party.33 In one sense, this is encouraging, since the majority logically bears greater responsibility for legislative action. On the other hand, this can also be viewed as somewhat troubling, since the minority party does possess tools to delay and obstruct the legislative process— particularly in the Senate—but the public apparently does not assign the minority party any significant responsibility for shortcomings in congressional productivity. Regardless of normative implications, this is a noteworthy finding because it helps to reject the alternative hypothesis that the public does not really distinguish between the parties when evaluating Congress. The second row of Table 1 shows that perceptions of an overly partisan Congress negatively affect public evaluations of both congressional parties. This is important because it shows that the public does assign the minority party at least some responsibility for the functioning of Congress. If it turns out that voters take their evaluations of the minority party into account when they vote, this would suggest that the minority party is held accountable for its contribution to the partisan tone in Congress. The final performance-based criterion is policy representativeness. An initial question is whether Americans actually have a realistic sense of where the two parties stand ideologically. In the sample, 96 percent of respondents were able to rate both parties. On a five-point scale running from liberal (1) to conservative (5), the average respondent rated congressional Democrats a 2.4 and congressional Republicans a 3.5—significantly more conservative (p<0.001).34 This finding is consistent with the argument that Americans are capable of distinguishing 33. Although I use the terms “majority party” and “minority party” for convenience, it is important to remember that respondents were not provided with this cue, nor were they told which party was in the majority. 34. The average respondent was located in between the two congressional parties, at 3.3. David R. Jones 481 A U TH O R C O PY between the two parties in Congress. More to the point, in Table 1 the absolute policy distance between a respondent and congressional Republicans has a negative effect only on evaluations of congressional Republicans. Likewise, the absolute policy distance between a respondent and congressional Democrats has a negative effect only on evaluations of congressional Democrats. For both referents, policy distance from the opposite party actually has a positive relationship on evaluations, though this effect is considerably less substantial. Overall, these findings support the idea that the public evaluates each party in Congress on the basis of its own policy stance, not on the basis of a single, overall impression of congressional policy. It is instructive to compare the effects of the variables on evaluations of each congressional party to their effects on evaluations of collective Congress. As expected, the effects of the performance-related variables on evaluations of the collective Congress share a much greater similarity to their effects on the majority party than on the minority party. This is particularly true regarding the productivity and policy representativeness criteria. This similarity helps to confirm that Americans associate the overall performance of Congress much more with the majority party than with the minority party.35 Nevertheless, despite this close connection between evaluations of the majority and of the collective Congress, they are far from identical. For example, the collective Congress is penalized heavily for the partisan tone of Congress, but this variable’s effect on the majority party is about 30 percent smaller (0.47 versus 0.34), perhaps because the minority party is sharing some of the blame. Also, the degree of policy divergence from the congressional majority, while important for evaluations of the collective Congress, plays a considerably larger role in evaluations of the majority party (0.12 versus 0.47). These latter observations support the notion that Americans’ impressions of Congress are not simply projected unthinkingly onto the congressional majority party. Rather, evaluations of the majority party in Congress are formed with that particular congressional party in mind. Finally, examining the probabilistic effects for each model reveals that the effects of the institutional performance-related variables (first four rows) are generally comparable to the effects of the other explanatory factors. For example, regarding evaluations of congressional Republicans, the influence of a respondent’s policy difference from Republicans in Congress is about the same as that of respondent’s partisanship, and the influence of Congress’s partisan tone and level of productivity are each about the same as that of the national economy. Regarding evaluations of congressional Democrats, the effect of a respondent’s 35. Additional analysis (not shown here) finds that when evaluations of the majority party and evaluations of the minority party are added as independent variables in model 1, only evaluations of the majority party have a significant effect on ratings of the collective Congress. 482 A MORE RESPONSIBLE TWO-PARTY SYSTEM? policy difference from Democrats in Congress is slightly larger than that of a respondent’s partisanship, and the effect of Congress’s partisan tone is nearly twice as large as the effect of a respondent’s feelings of external efficacy. In other words, public evaluations of the two congressional parties are not only statistically related to performance-based criteria, they are also strongly substantively related. Analysis of Electoral Accountability for Congressional Performance A U TH O R C O PY Having established that Americans do form distinct evaluations of the congressional majority party and the congressional minority party and that both majority and minority evaluations are influenced by performance-related criteria, I turn to the central empirical question in this study: do Americans hold both parties electorally accountable for their respective job performances in Congress? The theory section above hypothesized that the more positively a voter rates the performance of a given party in Congress, the more likely that voter will be to vote for candidates from that party, thus helping to hold the party electorally accountable. Accordingly, I test whether this expectation holds true for both parties in House elections. For the initial analysis of House elections, I continue to use survey data from the 2006 CES. In this case, however, the dependent variable captures whether a respondent intends to vote for the Republican candidate for the U.S. House as opposed to the Democratic candidate. The two key independent variables are citizens’ evaluations of Republicans in Congress and of Democrats in Congress. I expect that more positive evaluations of congressional Republicans will increase the likelihood of voting for the Republican candidate, and more positive evaluations of congressional Democrats will decrease that likelihood. To help rule out the possibility of spurious findings, the analysis controls for other major causal influences on House voting: a respondent’s ratings of the president’s performance in office, of a running incumbent’s performance in office, and of the national economy; recognition of the Republican candidate’s name and of the Democratic candidate’s name; and respondent ideology and party identification.36 Table 2 presents the results of a logistic regression analysis of House voting. Consistent with expectations, the results show that evaluations of Republicans in Congress have a positive and significant effect on voting for the Republican candidate for the U.S. House, and evaluations of the Democrats in Congress have a significant, negative effect. In other words, part of the reason Republicans lost the House in the 2006 elections was because negative assessments of their performance in Congress helped turn many voters against Republican candidates. 36. See the Appendix for details on coding. David R. Jones 483 Table 2 Explaining Voting for Republican House Candidates, 2006 Republican vote for House 0.976* −0.465* 0.281* 0.691* −0.836* 0.345 −0.031 0.245* −0.039 0.661* −2.167* (0.173) (0.154) (0.113) (0.187) (0.219) (0.291) (0.305) (0.148) (0.156) (0.085) (0.875) 0.74 −0.43 0.27 −0.59 0.68 0.09 −0.01 0.24 −0.04 0.75 PY Effect 0.796 R Note: Number of cases is 733. *p<0.05 (one-tailed). SE C O Rating of congressional Republicans Rating of congressional Democrats Rating of president Rating of incumbent (Republican) Rating of incumbent (Democrat) Recognize Republican candidate Recognize Democratic candidate National economy Ideology Party identification Constant Nagelkerke R2 Coefficient A U TH O In addition, voters who had positive evaluations of the performance of Democrats in Congress were more willing to perceive Democratic candidates as an acceptable alternative. To provide a sense of the relative size of these effects, the last column in Table 2 shows the estimated change in the probability of voting Republican as a voter goes from the lowest to the highest value of a variable, holding all other variables constant at their midpoint.37 Specifically, giving congressional Republicans the most positive rating as opposed to the most negative rating increases the probability of voting for the Republican House candidate by 0.74. Giving congressional Democrats the most positive rating as opposed to the most negative rating decreases the probability of voting for the Republican House candidate by 0.43. In comparison to many of the other effect sizes in the table, these effects are quite substantial. For example, they are in the same general range as the effect size for ratings of one’s own incumbent (0.59 to 0.68). Overall, these findings support the hypothesis that voters’ evaluations of both parties in Congress have an important impact on voting in House elections. 37. I use the midpoint because later analyses include multiple election years, and consistently picking the midpoint increases comparability across all of these models. 484 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY One potential concern with the results presented in Table 2 is that it is not possible to be certain from these data alone that the causal direction of the relationships is as hypothesized: current evaluations of congressional parties affecting subsequent vote preferences. It is possible, for example, that voters first make a decision about which candidate to support, and then rationalize their decision by rating that candidate’s party’s congressional performance more highly. In Table 2, all variables, including the dependent variable, are taken from the same, initial wave of the 2006 CES survey, making it difficult to determine which opinions came first. Demonstrating causal direction requires information about the same voters’ preferences over time. Fortunately, the 2006 CES included a second wave of interviews with the same respondents after the election. This “panel” format allows for a test of causal direction. First, I test a model in which the dependent variable is a voter’s final reported vote (wave 2), and all of the independent variables measure a respondent’s preelection opinions (wave 1). These independent variables are the same as those in Table 2, with one important exception. Crucially, the model adds a control for each respondent’s pre-election vote preference. Because this one variable is expected to have tremendous explanatory power on vote choice, this creates an extremely stringent test of my hypothesis. The only remaining variance in the dependent variable reflects subsequent changes in vote preference after the initial interview. Therefore, if evaluations of the majority and minority parties in Congress display any effects in this model, we can be confident they are affecting future vote preferences. Next, I test for evidence regarding the reverse causal story: that a preelection vote decision might be rationalized into subsequent positive feelings towards the chosen candidate’s party in Congress and negative feelings towards the rejected candidate’s party in Congress. In these two models, the dependent variables are, respectively, wave 2 evaluations of the Republicans in Congress and wave 2 evaluations of the Democrats in Congress. Wave 1 evaluations of each congressional party serve as controls. If wave 1 vote choice has a significant effect on subsequent evaluations of the congressional parties, this would support the reverse causal story. If not, this would be consistent with my causal argument. The results of these time-order tests are reported in Table 3. As expected, the results offer support for my causal argument, but none whatsoever for the reverse causal claim. In the first model, evaluations of congressional Republicans have a positive and significant effect on future voting for a Republican candidate, as expected. There is also some evidence (albeit only significant at p<0.10) that evaluations of congressional Democrats have the expected negative effect on future voting for a Republican candidate. In contrast, the second and third models show that vote preference clearly has no effect on future evaluations of congressional Republicans or congressional Democrats, respectively. While these results cannot be considered definitive, the evidence that is available suggests that the PY Table 3 Testing for Time-Order of Effects Rating of congressional Democratst SE Coefficient SE Coefficient SE (0.560) (0.308) (0.265) (0.196) (0.306) (0.327) (0.485) (0.488) (0.251) (0.288) (0.156) (1.707) 0.205 0.604* −0.119 0.474* 0.094 −0.218* −0.138 −0.206 0.032 0.116 0.145* (0.334) (0.131) (0.111) (0.112) (0.113) (0.128) (0.194) (0.201) (0.104) (0.095) (0.071) 0.217 −0.276* 0.768* −0.288* 0.090 0.149 0.544* 0.044 −0.092 −0.011 −0.325* (0.333) (0.130) (0.115) (0.110) (0.113) (0.129) (0.196) (0.201) (0.104) (0.094) (0.072) TH O A O C 2.429* 0.602* −0.420† 0.553* 0.819* −0.405 1.516* −1.215* 0.190 0.699* −0.010 −5.417* 0.857 U Vote preference for Houset−1 Rating of congressional Republicans t−1 Rating of congressional Democratst−1 Rating of presidentt−1 Rating of incumbent (Republican)t−1 Rating of incumbent (Democrat)t−1 Recognize Republican candidatet−1 Recognize Democratic candidatet−1 National economyt−1 Ideologyt−1 Party identificationt−1 Constant Nagelkerke R2 R Coefficient 0.548 0.545 Note: Number of cases is 447 in all models. The first model employs logistic regression. The second and third models employ ordinal regression (threshold terms are suppressed from the table to conserve space). *p<0.05, †p<0.10 (one-tailed). David R. Jones Rating of congressional Republicanst Republican vote for Houset 485 486 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY results found in Table 2 are not a result of reverse causality, but instead represent a true relationship, as hypothesized. While the results from Tables 2 and 3 provide strong evidence in support of this study’s arguments, it is reasonable to wonder whether the results hold true outside of the two specific elections I have analyzed. Replicating these analyses in different elections provides two potential benefits. First, it can increase confidence that the findings are generalizable across a variety of political contexts. Second, it can reveal possible conditional relationships that are not visible from analysis of a single election. In particular, are there any systematic differences in the size or significance of effects depending on which party controls each branch of government? To address these issues, I searched the iPoll database at the University of Connecticut’s Roper Center to find all election-year surveys that asked respondents to rate the job performance of Republicans in Congress and of Democrats in Congress.38 Using these search criteria, I was able to find appropriate surveys covering eight different House elections from 1994 through 2012.39 These elections cover a wide variety of political contexts, including unified Democratic control of government (1994, 2010), unified Republican control (2006), divided government with a Democratic president (1996, 2000), divided government with a Republican president (2008), split party control of Congress with a Republican president (2002) and a split Congress with a Democratic president (2012). As such, the data provide a strong test of the argument that individuals regularly hold both parties electorally accountable for their performance in Congress. For each of the eight years in which these surveys ask about an individual’s House vote, I estimate the same essential model of House voting used in Table 2, with a few notable exceptions. First, while the surveys all ask about the same core concepts, the exact question wordings vary slightly across them, meaning that that it is not safe to pool them into a single regression. Second, one of the surveys (1994) is missing a control for the economy, and all are missing questions about name recognition for House candidates. However, none of these variables is significant in the earlier analysis of House voting, so their absence is unlikely to affect the results and conclusions. Finally, a potentially more important exception is that none of the surveys contain the data necessary to control for performance evaluations of a voter’s running incumbent House member. To rule out the 38. In order to cast the widest possible net, I include four surveys that ask about the performance of each party’s “leaders in Congress” (2002, 2004, 2008, 2012) and one survey that asks about the performance of each party “in the U.S. House of Representatives” (1994). These differences in question wording do not appear to cause any systematic differences in the results. 39. In the few cases in which more than one appropriate survey was available, I selected the survey conducted closest to the election. In 2006, the ABC survey was closer to the election than the CES survey used in Table 2. In 1998 or 2004, no surveys asked about both congressional party performance and House vote. See the appendix for details on each survey. David R. Jones 487 Table 4 Effect of Performance in Office on Voting for U.S. House, 1994–2012 Voting for a candidate from the House majority party Effect of Effect of majority Effect of minority performance party’s party’s of president performance in performance in congress congress 0.35* −0.35* NA −0.26* 0.17* NA 0.49* −0.38* 0.35* −0.44* 0.35 PY −0.13 −0.40* NA −0.48* −0.23* NA −0.20* −0.56* −0.45* −0.41* 0.36 C O 0.20* 0.28* NA 0.38* 0.27* NA 0.39* 0.77* 0.27* 0.27* 0.35 R 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Average absolute effect TH O Note: In this table, majority party is defined by House control in split-control Congresses, but this particular decision does not affect the substantive findings either way. *p<0.05 (one-tailed). A U possibility that this omission might bias the results in favor of the study’s hypothesis, I reran the analysis of the CES data from Table 2 without controlling for this factor. The effect sizes for evaluations of congressional Republicans and evaluations of congressional Democrats were essentially unchanged, suggesting that the omission of this control variable is unlikely to bias the results in other election years either. Table 4 summarizes the results of the eight separate logistic regression analyses—one for each of the eight different House election years. In an effort to conserve space and not distract the reader with extraneous information, I present results only for three key independent variables: job performance of congressional Democrats, job performance of congressional Republicans, and (as a point of comparison) job performance of the president.40 The numbers in each cell represent the effect size of each variable, calculated in the same manner as in the final column of Table 2. 40. Details of each regression are available upon request. 488 A MORE RESPONSIBLE TWO-PARTY SYSTEM? Discussion TH O R C O PY The results are organized so as to compare the effects of majority party performance ratings and minority party performance ratings on voting for a candidate from the House majority party.41 Significant effects are noted with asterisks and the bottom of the table tallies the average absolute effects across all elections. There are several notable findings in these eight House elections. First, out of the sixteen coefficients that were estimated for majority and minority party performance, fifteen are statistically significant. This suggests that the results found in Table 2 are not unique to 2006, but instead can be found across a wide range of different electoral contexts. Second, the averages at the bottom of the table indicate that not only are the electoral effects of the minority party’s performance ratings statistically significant, they are also roughly the same size as those for the majority party (0.36 and 0.35, respectively). Third, in addition to being roughly equal to one another, the size of the effects of majority party evaluations and minority party evaluations are also relatively substantial: each is about as large as the effect of presidential performance—a variable that is commonly cited as a powerful determinant of House voting.42 Taken together, these findings provide an important corrective to the literature on congressional accountability in House elections. Existing research implies that it is primarily the majority party that is held accountable for its performance in Congress. In contrast, the results in Table 4 show that evaluations of both majority and minority party performance in Congress regularly and substantially affect voting in House elections. A U This study began by asking whether the minority party has any electoral incentive to act responsibly in Congress. The short answer is: yes. Americans do hold meaningfully different evaluations of each party’s performance in Congress. To a significant degree, these evaluations are formed on the basis of criteria logically related to legislative performance. In turn, Americans use the evaluations of both parties when casting their votes in House elections. Importantly, the electoral impact of minority party performance ratings appears to be just as large, on average, as the impact of majority party performance ratings. Furthermore, the magnitude of these effects is substantial. It is comparable, in fact, to other prominent predictors of vote choice. In sum, accountability for congressional performance is not simply a referendum on the party in power. Instead, 41. For convenience, in the two election years in which party control of Congress was split by chamber (2002, 2012), I treat House control as majority control. This decision does not affect the overall pattern of results. 42. For example, see Jacobson, Politics of Congressional Elections. David R. Jones 489 A U TH O R C O PY congressional accountability is actually closer to the “responsible two-party system” ideal set by the 1950 APSA report more than six decades ago. One important consequence of these findings is that considering the evaluations of the minority party as well as the majority party could help improve election forecasting, especially surrounding periods of split party control of Congress. Under the existing one-party referendum model of congressional accountability, forecasters can use survey data on public evaluations of Congress to help predict vote and seat losses in Congress, but theoretically this formula only applies to the party that “controls Congress”—a concept made ambiguous by split-party control.43 This study’s two-party model of congressional accountability provides the potential basis for a forecasting formula that can be applied to both chambers regardless of which party is in control of each. As survey organizations begin to ask these party-specific congressional performance questions on a more regular basis—as they appear to be doing since 1994—these data have the potential to improve our ability to both explain and predict aggregate election outcomes. Another implication of this study is that the threat of electoral losses creates an incentive for party leaders to attempt to rein in unpopular behavior on the part of their members in Congress. This idea may help explain recent changes in the strategies of Republican leaders (though not necessarily in the behavior of the rank and file). In the first two years after the 2010 election, Republican leaders seemed content with their party in Congress continuing to eschew bipartisanship, vote in a highly ideologically conservative manner,44 and block action on major legislative agenda items. But since witnessing their party’s failure to recapture the White House and its loss of seats in both the House and Senate in 2012, Republican Party leaders appear much more intent on reining in their contingent in Congress. The Chair of the Republican National Committee published a report “push[ing] for a tone of ‘tolerance and respect’ in the immigration debate.”45 John Boehner has helped facilitate the passage of more bipartisan bills in Congress.46 Eric Cantor has tried to refocus his party caucus on a less ideological, more practical legislative agenda.47 While ultimately these efforts may not prove successful, the fact that they are being made suggests that party leaders are aware that what the party does in Congress affects its future electoral prospects. 43. John Sides, “How Congressional Dysfunction Could Hurt House Republicans,” Washington Post, January 13, 2013. 44. Nolan McCarty, Keith Poole, and Howard Rosenthal, Polarized America. 45. Steve Peoples, “Republican Roadmap Calls for Immigration Overhaul,” Boston Globe, March 19, 2013. 46. Jonathan Chait, “How Boehner’s Crazy Method Is Letting Obama’s Agenda Move,” New York Magazine, April 11, 2013. 47. David M. Drucker, “Cantor Moving Legislation Aimed at Rehab[b]ing GOP Brand,” Roll Call, March 11, 2013. 490 A MORE RESPONSIBLE TWO-PARTY SYSTEM? A U TH O R C O PY Of course, just because a party has electoral incentives to act responsibly in Congress does not mean that these incentives alone will always be determinative of congressional behavior. These incentives may sometimes be counterbalanced—or even overwhelmed—by other incentives to act irresponsibly. This may be particularly true for the minority party during unified government. For example, in 2009–2010 Democrats held the presidency and a majority in both chambers of Congress. As such, they could expect the public to hold them fully responsible for the performance of the president and at least partly responsible for the performance of Congress. The minority Republican Party, in contrast, could anticipate that it would be held only partly responsible for what went on in Congress, and not at all responsible for the performance of the president. Therefore, Republicans might rationally calculate that while a strategy of partisan obstruction could slightly harm their party by lowering its performance rating in Congress, they could more than make up for this slight harm by lowering the performance rating of Democrats in Congress and of President Obama, thus improving their electoral prospects on balance.48 Finally, this study helps highlight the fact that party polarization in Congress has been a mixed blessing for American politics. On the one hand, polarization has helped clarify the distinct roles that each party plays in Congress. In turn, this has made it possible for Americans to apply electoral accountability to both congressional parties, as opposed to just the president’s party. On the other hand, the fact that voters now appear more aware about what the parties are doing in Congress—and more willing to use that information when casting ballots—may have increased the incentives for parties to polarize even further. In particular, parties may now view the legislative branch less in terms of a venue for solving public policy problems, and more in terms of a venue for maximizing partisan electoral advantage. David R. Jones is Professor of Political Science at Baruch College and the Graduate Center, City University of New York. His research on congressional behavior and elections has been published in several scholarly journals. He is also the author of Political Parties and Policy Gridlock in American Government (2001) and coauthor, with Monika L. McDermott, of Americans, Congress, and Democratic Responsiveness: Public Evaluations of Congress and Electoral Consequences (2009). 48. According to Table 4, the combined negative effects of these variables on Democratic electoral prospects would likely be greater than the negative effect of congressional Republican performance on Republican electoral prospects. David R. Jones 491 Appendix Data Used in Tables 1–3 A U TH O R C O PY Coding: Rating of Congress: scores range from −2 (strongly disapprove) to +2 (strongly approve); Rating of congressional Democrats: scores range from −2 (most unfavorable) to +2 (most favorable); Rating of congressional Democrats: scores range from −2 (most unfavorable) to +2 (most favorable);49 Unproductive Congress: scores range from 1 (high grade for tackling key issues) to 5 (low grade for tackling key issues); Overly partisan Congress: scores range from 1 (high grade for bipartisanship) to 5 (low grade for bipartisanship); Policy divergence of Republicans: absolute difference between respondent’s rating of self and of congressional Republicans on a 5-point ideological scale, with scores ranging from 0 to 4; Policy divergence of Democrats: absolute difference between respondent’s rating of self and of congressional Democrats on a 5-point ideological scale, with scores ranging from 0 to 4;50 National economy: scores range from 1 (much worse) to 5 (much better); External efficacy: scores range from 1 (strongly agree that officials don’t care about people like me) to 5 (strongly disagree that officials don’t care about people like me); Party identification: self-reported party identification including leaners, coded as 1 = Democrat, 0 = independent, 1 = Republican; House vote: dummy variable coded 1 for the Republican candidate and 0 for the Democratic candidate; Rating of president: scores range from −2 (strongly disapprove) to +2 (strongly approve); Rating of incumbent: respondent’s rating of their own, running incumbent on a scale from −2 to +2, with open seat races coded 0; Recognize Republican candidate: dummy variable coded 1 if respondent recognized the Republican candidate’s name out of a list of four names, 0 otherwise; Recognize Democratic candidate: dummy variable coded 1 if respondent recognized the Democratic candidate’s name out of a list of four names, 0 otherwise; Ideology: respondent’s self-placement on a scale from 1 (very liberal) to 5 (very conservative). Source: 2006 Congressional Elections Study conducted by the Center for Survey Research at Indiana University, September 28-November 6, 2006, N = 1,023 (wave 1) and November 9-December 30, 2006 (wave 2; 75 percent 49. Ideally, I would have preferred to use a traditional approval/disapproval question for each congressional party, but these were not asked by the CES. Fortunately, evidence suggests that in practice the distinction between approval and favorability is very small: in the same survey, presidential approval correlated with presidential favorability at p = 0.86. 50. To guard against the possibility that respondents have a biased perception of the ideology of each congressional party, I also ran the models substituting a congressional party’s sample-wide mean rating in place of each individual’s rating in this formula. The results are substantively similar. Because the study’s expectation is that individuals’ perceptions are what matter, the tables report respondents’ own ratings rather than the sample mean. 492 A MORE RESPONSIBLE TWO-PARTY SYSTEM? recontact rate). Note: survey uses telephone interviews with a sample of adults from the forty-eight contiguous states. Data Used in Table 4 A U TH O R C O PY Coding: House vote: dummy variable coded 1 for the Republican candidate and 0 for the Democratic candidate; Rating of congressional Republicans: respondent’s rating of the job performance of Republicans in Congress, coded as −1 = disapprove, 0 = neutral, 1 = approve; Rating of congressional Democrats: respondent’s rating of the job performance of Democrats in Congress, coded as –1 = disapprove, 0 = neutral, 1 = approve; Rating of president: respondent’s rating of the president’s job performance, coded as −1 = disapprove, 0 = neutral, 1 = approve; National economy: scores range from 1 (very poor) to 4 (very good); Ideology: respondent’s self-placement on a scale, coded as −1 = liberal, 0 = moderate, 1 = conservative; Party identification: self-reported party identification including leaners, coded as 1 = Democrat, 0 = independent, 1 = Republican. Sources: ABC News/The Washington Post Poll: Politics, conducted by Chilton Research Services, October 27–31, 1994, N = 1,015; NBC News/Wall Street Journal Poll: 1996 Election/ Economic Issues, conducted by Hart-Teeter Research Companies, May 10–14, 1996, N = 1,001; Gallup/CNN/U.S.A. Today Poll # 2000-35: Post-Democratic Convention Poll, conducted by Gallup Organization, August 18–19, 2000, N = 1,043; PSRA/ Pew Poll #06: June News Interest Index—Politics/First Ladies/Bush/Mideast /Terrorism, conducted by Princeton Survey Research Associates, June 19–23, 2002, N = 1,212; ABC News/Washington Post Poll # 2006-1023: Politics/Congressional Elections/War in Iraq/Ethics, conducted by ABC News and The Washington Post, October 5–8, 2006, N = 1,204; CNN/ORC Poll # 2008-016: 2008 Presidential Election/Financial Crisis, conducted by Opinion Research Corporation, October 3–5, 2008, N = 1,006; ABC News/Washington Post Poll: September Monthly— Congressional Elections, conducted by ABC News and The Washington Post, September 30–October 3, 2010, N = 1,002; CNN/ORC Poll: 2012 Presidential Election, conducted by Opinion Research Corporation, November 2–4, 2012, N = 1,010. Note: All coding follows the same basic scheme presented above, with exceptions noted in the text. All surveys use telephone interviews with a national sample of adults. All analyses consist of registered and/or likely voters only.
© Copyright 2025 Paperzz